general system theory.pdf
TRANSCRIPT
-
8/19/2019 General System Theory.pdf
1/197
-
8/19/2019 General System Theory.pdf
2/197
.- rb. H)
¡SEMA
Service
INFOF2MATION
SCIENTIFIQUE
-
8/19/2019 General System Theory.pdf
3/197
VIEWS ON GENERAL
SYSTEMS THEORY
-
8/19/2019 General System Theory.pdf
4/197
PUBLICATIONS FROM THE SYSTEMS RESEARCHCENTER,
CASE INSTITUTEOF TECHNOLOGY
Ellis A. Johnson, Director
VIEWS ON GENERAL SYSTEMS THEORY
edited by Mihajlo D. Mesarovié
SYSTEMS: RESEARCH AND DESIGN
edited by Donald P. Eckman
-
8/19/2019 General System Theory.pdf
5/197
VIEWS ON GENERALSYSTEMS THEORY
proceedings of The Second Systems
Symposium at CaseInstitute
of Technology
editedbedited by , 0 L L)MIHAJLOD. MESAROVIC
Associate Professor of EngineeringSystems Research CenterCase Institute of Technology
John Wiley & Sons, Inc., New York London -Sydney
-
8/19/2019 General System Theory.pdf
6/197
Copyright © 1964 by John Wiley & Sons, Inc.
All Rights Reserved. This book or any part
thereof must not be reproduced in any formwithout the written permission of the publisher.
Librury of Congress Catalog Card Number: 64-14992
Printed in the United States of America
-
8/19/2019 General System Theory.pdf
7/197
CONTRIBUTORS
Russell L. AckofiProfessor of Operations Research, Case Institute of Technology,
Cleveland, OhioW. RossAshbyProfessor of Electrical Engineering, University of Illinois, Urbana,IllinoisKenneth E. BouldingProfessor of Economics, University of Michigan, Ann Arbor,MichiganAbraham Charnes
Research Professor of Applied Mathematics and Economics, North-western University, Evanston, Illinois
C. West ChurchmanProfessor of Business Administration, University of California,Berkeley, CaliforniaWilliam W. CooperProfessor of Economics and Industrial Administration, Carnegie Insti-tute of Technology, Schenley Park, Pittsburgh 13, PennsylvaniaRudolf F. DrenickProfessor of Engineering, Polytechnic Institute of Brooklyn, Brooklyn,New YorkR.W. GerardProfessor of Biological Sciences, University of California, Irvine,CaliforniaRobert Kalaba
'
StaffMathematician,
The RANDCorporation,
SantaMonica,California
William K. LinvillProfessor of Engineering, Stanford University, Stanford, California
Mihajlo D. MesaroviéAssociate Professor of Engineering, Case Institute of Technology,Cleveland 6, Ohio
v
-
8/19/2019 General System Theory.pdf
8/197
Vi CONTRIBUTORS
John MyhillProfessor of Philosophy, Institute for Advanced Study, Princeton,New Jersey
Hilary Putnam , .....Professor of the Philosophy of Science, Massachusetts Institute ofTechnology, Cambridge, Massachusetts
Anatol RapoportProfessor of Mathematical of Biology, University of Michigan, Ann
Arbor, MichiganLotfiA. ZadehProfessor of Electrical Engineering, University of California, Berkeley,California
-
8/19/2019 General System Theory.pdf
9/197
Dedicated to the Memory ofProfessor Donald P. Eckman
1915-1962Founder and First Director
of the Systems Research Center
-
8/19/2019 General System Theory.pdf
10/197
-
8/19/2019 General System Theory.pdf
11/197
FOREWORD*
It is a great pleasure to have this opportunity to meet the participantsof the Second Systems Symposium. A number of you also took part in
the first conference held two years ago. In the intervening years thedevelopment of a logical framework for the design of large and complexsystems has become of ever more pressing urgency.
One of the complicating problems is the lack of public understandingof the meaning of the term "system," as we use it, and of the goals of
systems research. The term is widely used by all segments of the popu-lation, but its meaning is clouded.
As a result, a request for support of a systems research project bringsdiverse reactions.
Yet one thing seems to be very much in our favor. There is generalagreement that the most crucial problems our society faces involve
large and complex systems.We cannot ignore these problems in the hope that they will go away,
or choose from those confronting us the ones that are simplest or easiestto solve. It is true that a chemical company can choose not to developa new process, but the alternative may be failure to keep abreast of
competition resultingin the eventual decline of the business. A nation
may elect not to create a defense system or modify its economic systemto meet the competitive pressures of the modern world, but again, thealternative may be disaster.
Strong arguments for serious and large-scale efforts on many frontsto develop a logical framework for the design of complicated systemscan be made, and there is clear and ample evidence to support the thesisthat better solutions to some of these large-scale systems problems cannow be found.
It is clear that the problem is interdisciplinary in character. There-fore the solution cannot be sought and found within the scope of anyone traditional area of activity but requires the contributions and
cooperation of all concerned.
*At the banquet,SecondSystemsSymposium,April1963.ix
-
8/19/2019 General System Theory.pdf
12/197
x FOREWORD
It is obvious that the most powerful available mathematical tools areessential, and that many of the problems are so complex that largecomputing machines have to be used. Wherever possible, dynamic
measurements must be made under actual operating conditions andused to help develop general theoretical formulations.In man-machine systems, biological understanding at both the micro
and macro levels is essential. 1 am speaking of a man-machine systemin which man or a subsystem of man appears as a component of thesystem. There are systems problems, again man-machine problems,where men appear in groups in the system. In this kind of system, nological framework of design can be formulated without a deeper under-
standingthan we now have of
group dynamics,of relations between
groups, and of relations between individuals and groups.The idea that it is possible to deal with integrated and intricate
man-machine systems through a general systems theory is interestingenough to attract this distinguished group of conference participants,but we must not forget our limitations. We still await a Clerk Maxwellto provide a complete and unifying theory; and if we are overzealousand attempt to expand our definition of a general systems theory toinclude the interrelation of all things, 1 do not believe we will achieveour goal. However, I believe we can achieve a framework of logicwhich will enable us to design and predict systems of far greater com-plexity than those which can be dealt with today.
1 do not believe that this goal can be achieved by limited sorties atthe existing front of knowledge. I am inclined to think that the timehas come to take bold action and risk landings well beyond the existingfrontier. The risk of such landings may be great, but the possiblerewards justify the risk.
Despite the sobering nature of these problems, a complex sociological-economic-engineering system, such as the predicted Great Lakes mega-lopolis which may develop through the merging of existing cities in theregion as our population grows, offers challenging experimental pos-sibilities. The introduction of planned and monitored diversity offersan exciting alternative to the more self-contained experiments of thecontrollable university research laboratory.
The image of our subject that I have, then, is that general systems
theory will be used as one of the major tools guiding the design of largeinterdisciplinary systems. Furthermore, such theory must itself bedeveloped with interdisciplinary support rather than solely by a specialgroup of new general systems theorists. The over-all cooperationneeded to insure the success of systems research will involve an in-terdisciplinary research team; participation and exchange of faculty
-
8/19/2019 General System Theory.pdf
13/197
FOREWORD xi
between several institutions; support by a number of sponsors, eachinterested in some particular aspect or subsystem of the over-all sys-tems problem; and a broader experimental base. The result of sucha
deliberately designedthree-dimensional
disciplinary-research agency-sponsored matrix might well have a revolutionary impact not only onthe rate of progress in systems research but also on the affairs of man.
JOHN HRONES
Vice President of Academic AffairsCase Institute of Technology
Cleveland,OhioFebruary
-
8/19/2019 General System Theory.pdf
14/197
-
8/19/2019 General System Theory.pdf
15/197
PREFACE
The last decade has witnessed a renewal of a strong interest amongscientists and engineers in a better understanding of the basic theoreti-cal concepts used in the various disciplines. Questions have been raised
as to whether the similarities among the basic concepts are sufficient to justify development of a more general theory which will represent abasis for the more specialized theories. The essential idea for such anundertaking is very appealing in its simplicity. Knowledge in anyparticular discipline is embodied in a theory which as such is an abstractconstruct. It is only natural to expect that many of these constructshave strong similarities or, perhaps, are essentially the same. It is, ofcourse, quite another matter actually to find these similarities as well
as new concepts which will bring diverse developments together. Whatmight be the road leading toward such a development, and what are the
prospects for its success, was the theme of the Second Systems Sym-posium held at Case Institute of Technology in the spring of 1963.
It has been pointed out in several circles that the starting point fora general theory can be found in the notion of a system or, rather, a
general system. The theory then becomes general systems theory. Iwill not venture here to elaborate on the meaning of such a theory.
After all, this is the subject of all the contributions in this volume.Let me only very briefly emphasize the motivation for developing sucha theory. Recently, it was very pointedly emphasized by Dr. SimonRamo that the human community is in a race: "Systems engineeringversus the rapidly increasing complexity of our growing technologicalcivilization." Systems engineering should offer some facilities tounderstand, and to control such complexities. In a broader context,the ever increasing complexity of the world we live in requires thatour
understandingof this
phenomenonbe based on a broader theoreti-
cal basis.The motivation for an attempt to develop such a theory is therefore
both scientific, with the objective of improving our understanding ofthe natural and social phenomena, and practical (engineering), withthe objective to provide better methods for the synthesis and controlof complex systems.
xiii
-
8/19/2019 General System Theory.pdf
16/197
xiv PREFACE
The idea of developing a general theory that would provide a moreunified picture of our experience is not a new one. Actually, this ideahas been very much cherished by classical thinkers and philosophers
and has been revived occasionally throughout history. What character-izes the present attempt toward such a theory, however, is its concernwith scientific rather than philosophical matters which, actually, are
being developed by scientists and engineers. The interest in the theoryarises from the needs of the practicing scientist and engineer ratherthan from scientific or philosophical curiosity. However, this is not to
imply that the theory might not have philosophical implications.The contributions comprising this book represent slightly revised
versions of the talks delivered at the Second Systems Symposium. Atthe time when the symposium was held, the objective of the symposiumwas stated to be the clarification of the following problems:
1. Basic characteristics of general systems theory.2. Review of major developments achieved to date.3. Statement of current problems of importance.4. Prospect for the future.
Seventeen speakers participated in the formal presentation and paneldiscussion, and more than 200 attendants engaged in active discussionduring the two days of the conference. At the time when the book wasscheduled to go into print, 14 manuscripts had reached the editor.
The manuscripts represent rather successfully the spectrum ofopinions expressed at the conference. All contributors agreed on one
point-necessity for the development of a general systems theory.However, opinions widely differed regarding the type of the theory
and the direction in which it should be developed. To appreciate allthe nuances, it is certainly advisable to go through all the contributions.
However, the editor had the task of trying to indicate the relations anddiversities that appeared in the discussion. In this particular instancethe editor's job was made more complex by the fact that he took anactive part in the discussion and provided a contribution to the pro-ceedings. 1 will restrict my comments, therefore, to the approach takentoward the problem of how to develop the general systems theory.
First ofall,
some of theparticipants
took a definitestand, venturingto define a system and then discussing the consequences of such a
definition. A second group of participants argued that the generalsystems theory should not be formalized since this very act will limitits generating power and make it more or less specific. A third groupproposed to consider systems theory as a viewpoint taken when one
approaches the solution of a given (practical) problem. Finally, it was
-
8/19/2019 General System Theory.pdf
17/197
PREFACE XV
expressed that a broad-enough collection of powerful methods for the
synthesis (design) of systems of diverse kinds should be considered as
constituting the sought-for theory and any further integration was
unnecessary. There were also participants that shared the viewpointsof more than one of the above groups.It is not apparent which of the viewpoints will prevail. It might very
well happen that several developments along different lines will takeplace. At any rate, it is believed that diverse viewpoints are presentedin the 14 contributions and, as such, they can present the cornerstonefor further developments of the general systems theory in whateverdirection appears to be promising.
Many persons have contributed their time and effort to make theSecond Systems Symposium possible. Dr. Simon Ramo, an old sup-porter and friend of the Systems Research Center, has helped with thesolution of the financial problems in connection with the conference.Dr. John Hrones, Vice-President of Case Institute, again has helpedthe Systems Research Center with the organizational problem as hehas done so many times in the past. Dr. Raymond Nelson, Directorof the Computing Center at Case Institute, was acting-director at theSystems Center at the time of the organization of the conference and his
enthusiastic backing of the project made it a success. Dr. Ellis John-son, present Director of the Systems Center, continued to support theproject in the same spirit until its final completion. Dr. Ray Bolz,Head of the Engineering Division at Case, served as the chairman ofthe session at the symposium. He has always been instrumental instimulating the systems activities in the Engineering Division.
The reader of the book should join the editor in acknowledgmentof the help of Dr. Lester Goodman, Assistant Director of the Systems
Research Center, who served as cochairman of the conference. Theresearch on present system theory in the Systems Research Center hasbeen supported by ONR under Contract 1141 (12). The results of thisresearch are reported in the Case contribution to this book. In par-ticular, Mr. R. Wilcox and Dr. M. Shelly of that office encourageddevelopment of that contribution through repeated visits to Case anddiscussions on the subject.
Finally a word seems appropriate regarding the arrangement and
sequence of the chapters. Since all the contributions represent, inessence, independent viewpoints, no grouping or proper sequencing ofthe chapters was attempted. Rather it was decided to put them inorder of their presentation at the conference. The reader can start withany chapter he chooses without any loss of continuity.
While listening to the presentations, Professor Boulding was inspired
-
8/19/2019 General System Theory.pdf
18/197
XV1 PREFACE
to write small poems about some of the contributions. We decided toinclude them, hoping that they might enlighten the reading and, in
themselves, present some kind of bridge between two cultures.
MIHAJLO D. MESAROVIÉCleveland,OhioFebruary 1964
-
8/19/2019 General System Theory.pdf
19/197
CONTENTS
1. Foundations for a General Systems Theory, Mihajlo D.Mesarovié 1
2. General Systems as a Point of View, Kenneth E. Boulding 25
3. The Concept of State in System Theory, L. A. Zadeh 39
4. General System Theory and Systems Research: ContrastingConceptions of Systems Science, Russell L. Ackoff 51
5. Constrained Extremization Models and Their Use in Devel-oping System Measures, Abraham Charnes and William W.
Cooper61
6. The Compleat Conversationalist: A "Systems Approach" tothe Philosophy of Language, Hilary Putnam 89
7. The Abstract Theory of Self-Reproduction, John Myhill 106
8. Entitation, Animorgs, and Other Systems, R. W. Gerard 119
9. Toward Approximate Analyses of Linear Dynamic Systems,William K. Linvill 125
10. Invariant Imbedding and the Analysis of Processes, RobertKalaba 143
11. System Causality and Signal Predictability, Rudolf F.Drenick 157
12. Introductory Remarks at Panel Discussion, W. Ross Ashby 165
13. Remarks on General Systems Theory, Anatol Rapoport 170
14. An Approach to General Systems Theory, C. West Churchman 173
Index 177
-
8/19/2019 General System Theory.pdf
20/197
-
8/19/2019 General System Theory.pdf
21/197
CHAPTER 1
FOUNDATIONS FOR AGENERAL SYSTEMS THEORY
MIHAJLO D. MESAROVIC
According to MesaroviéA set of proper statements which
Has mastered, in well-ordered schoolsA set of transformation rulesWhich rules in turn have rules to twist 'emDeserves the name of general system.
All systems, it is now proposedAre either open, or are closed,The closed have one-to-one relationsBut don't result in innovations.The open are disturbed, adaptiveOr Heisenberg-observer-captive.
K. B.
INTRODUCTION
Since the very beginning of scientific inquiry, emphasis has alter-
nated between the particular (data and facts) and the general(similarities and théories). Theory requires the facts on which to
build, and facts and evidences need theory for understanding of the
phenomena. Traditionally, theories have been developed within theboundaries of a particular field; i.e., they have been concerned with a
given class of real systems such as biological, physical, chemical, etc.
Rapid advances in the collection of data, as well as the developmentof theories within particular fields of scientific inquiry, have created
a new problem on a higher level. Namely, a need appeared for a gen-eral theory which would serve as a foundation for other specifictheories, crossing the boundaries of the different fields, and ultimatelyresulting in a deeper understanding of the world we live in.
In addition to the scientific interest, the solution of certain very
practical problems appears to require the help of such a general theory;1
-
8/19/2019 General System Theory.pdf
22/197
2 VIEWS ON GENERAL SYSTEMS THEORY
hence development of this theory has become a matter of paramountimportance. For example, in engineering (taken in a very broad sense),a general theory that crosses traditional fields appears necessary for
the design of systems that include men, computers, and machines.Similarly, in the sciences, the proper understanding of certain phenom-ena, or certain real systems, requires their observation in the broadercontext of larger systems which now include the systems studied tradi-
tionally by the different fields.In this chapter, a definite attempt is made to lay down the foundation
for the development of such a general theory. Before elaboratingfurther, the basic motivation and justification for the selection of a
theory of this type are discussed. Only the basic concepts of thetheory are outlined here because of space limitations. Sufficient de-tails are presented, however, to indicate the approach taken. A fulleraccount of the theory will be presented in a forthcoming book."
MOTIVATION
As mentioned in the foregoing section, both scientific and tech-nological advances exhibit a need for a general theory. Let us con-sider the motivations for such a theory.
A basic methodological tool used in the development of scientifictheories is the method of analogy. It suffices here to discuss thismethod informally. In essence, the method of analogy has the follow-
ing underlining philosophy: An equivalence relation is established be-tween two systems (either of the systems may be real or conceptual).
The equivalence relation indicates that aspects of the behavior ofthe two systems which are the same. One can then make certain in-ferences about the behavior (or perhaps some relevant aspect of it)of one system from the study and observation of the other. The se-lection of the system to be studied is based on practical consideration;namely, the system on which the observations are to be made isselected so as to be simpler than the actual system in all respects notrelated by the established equivalence. The system used in the in-
vestigation is then called a model.It is interesting to note that in science and in mathematics some-
what opposite directions are followed in the selection of a model. Inthe sciences, the model is invariably an abstract, mathematical sys-tem, or a much simpler physical system than the original. In mathe-
matics, however, a model is a more specific and less abstract analogy
-
8/19/2019 General System Theory.pdf
23/197
FOUNDATIONSFORA GENERALSYSTEMSTHEORY 3
of the actual system. In the theory proposed here the model can beunderstood in either sense.
Since the theory of any specific scientific field depends on analogies,
it seems reasonable that the general theory should similarly rely on ageneral type of analogy. Also, the models used in the general theoryshould be of a general kind.
We turn now to engineering. The building of complex systems,particularly those involving men, posed new problems which couldnot be solved with the traditional tools and techniques. A new field
consequently developed under the name of Systems Engineering. Sofar it has relied, to a large extent, on the tools and techniques of other
fields,combined with the
practitioner's experienceand intuition. The
principal reason for this was a lack of theoretical foundations fromwhich methodological procedures could be derived. One of the ob-
jectives of general systems theory is to provide such a foundation.
REQUIREMENTS
We consider now the basic requirements that a general theoryshould satisfy in the light of the motivations discussed in the fore-going section. The following two appear to be among the most im-
portant.
1. The general theory should be general enough to encompass dif-ferent types of already existing specific theories. It should, therefore,be sufficiently abstract so that its terms and concepts are relevant to
specialized theories. Clearly, the more abstract statements have a
broader context but, at the same time, they carry less information re-garding the behavior of any particular system. The general conceptsmust emphasize the common features of all the systems considered yetneglect the specific aspects of the behavior of any particular system.The real challenge in developing a general theory is, therefore, to findthe proper level of abstraction. The concepts must have wide ap-plication, while the conclusions which they lead to must provide suf-ficient information for proper understanding of the particular class of
phenomena under consideration.2. The general theory has to have a scientific character in the sensethat its concepts and terms must be uniquely defined within the propercontext. If the general theory is to be of any help in solving scientificand engineering problems, it must not rely on vague, ill-defined, almost
poetic analogies. The basis for the general theory must be solid so
-
8/19/2019 General System Theory.pdf
24/197
4 VIEWS ON GENERAL SYSTEMS THEORY
that its conclusions have practical meaning for real systems. Relianceon vague analogies is the main weakness of the general theories pre-viously conceived. Let us hasten, however, to add that by requiring
scientific status for the general theory we do not want to imply thatit should somehow consist of the set of dogmas revealing the essentialtruth of the real phenomena. On the contrary it is to be expected thata general theory, much as other human intellectual innovations, will
age, be updated and rejuvenated. However, at any given time,within the framework of existing knowledge, the general theory shouldbe designed on solid foundations.
In the following sections a general theory will be proposed which
is abstract enough yet definite enough to be useful. In the spirit ofthe remarks in this section, however, none of the basic concepts ofthe theory should be considered as unalterable. Only the type of the
proposed theory and its spirit are to be considered as invariant.
BASICCHARACTERISTICSOF THE THEORY
The general theory proposed here has the following basic character-istics :
1. It is developed around the notion of a system. Since it can be
argued that both science and engineering are concerned with the studyof real systems and their behavior, it follows that a general theoryshould be concerned with the study of the general systems. In thenext section, the concepts of general systems will be introduced on atechnical basis. It suffices for the present discussion to consider a
general system as an abstract analog or model of a class'of real sys-tems. General systems theory is then a theory of general models.It might be argued that a general theory should be broader than the
concept of systems, but, in order to preserve the definite character ofthe theory as discussed in the foregoing section, it appears to be ofconsiderable advantage to base the theory on the concept of a system.
2. General systems theory, as a theory of general models, must en-compass the specific theories concerned with the more restrictive types
of models; e.g., the theory of linear systems, the theory of Markovsystems, etc. These theories can be regarded as being concerned withthe specific types of models. The delineation between general systemstheory and the theory of specific systems is very difficult and oftenrather arbitrary. This is, however, as it should be since, in principle,there is nothing to be gained from constructing sharp boundaries
-
8/19/2019 General System Theory.pdf
25/197
FOUNDATIONSFORA GENERALSYSTEMSTHEORY 5
among the fields. Flexibility in this respect will only help further
development and growth. However, this should not be confused withthe requirements for solid foundations as argued in the last section.
No matter how we delineate the field, the basic concepts and notionsmust be consistent with the context proposed.3. General systems theory also unifies the theories of different
aspects of systems behavior such as communication, control, adapta-tion, learning self-organization, theory of computing and algorithms,etc. This unification is, of course, on a more abstract level. It also
helps someone interested in a specific aspect of the systems behaviorto understand this in a broader context.
4. Generalsystems theory
in itsdevelopment
useslargely
moreabstract branches of mathematics and in this sense is related to themathematical theory of formal systems. However, the meaning of theterms used is of primary importance to general systems theory whileit is largely immaterial for formal mathematics. This becomes espe-cially clear when discussing the goal-seeking behavior of general sys-tems. For example, from a general systems viewpoint, the solution ofan equation might be important only insofar as it reveals the behaviorof the system under consideration.
5. We are now in a position to outline more clearly the objectivesand scope of general systems theory and to consider the basicmethodology used. General systems theory is a scientific undertakingconcerned with the study of the behavior of general systems. Itsmethodology is essentially scientific in nature. The relation betweenthe investigator and the system under study is analogous to the rela-tion between nature and the natural scientist. Namely, a generalsystem is given either in terms of mathematical equations or com-
puter specifications, or a physical analog, or a verbal description, etc.,and the objective is to discover the basic properties of its behavior.
CONCEPT OF A GENERALSYSTEM
The problem of selecting the proper level of generality for the basisof general systems theory is indeed formidable. We shall somewhat
circumvent this problem, in this section, by introducing the conceptof a system in the most general context while using more specifictypes of systems in the development of the theory.
The concept of the general system introduced in this section is in-structive since it indicates how one can discuss the behavior of a real
system by using only linguistic constructions while simultaneously
-
8/19/2019 General System Theory.pdf
26/197
6 VIEWS ON GENERAL SYSTEMS THEORY
satisfying the requirements of a definite basis, as discussed in the
foregoing section.
Three different definitions of a general system are introduced, but
all three will be on the same level of abstraction.
IlThe first and most general approach is to use a linguistic definition
/ for a general system. To arrive at this concept of a general system,
several preliminary concepts must first be introduced.
One starts with the concept of a formal statement in a language L.
The language under consideration may be a natural language as, e.g.,
English, a computer language, or any formal language generated in
a recursive manner by a machine. A formal statement F in a language
L is defined as a sentence which is formed according to the rules ofgrammar of the respective language but the truth of which is not
revealed by the statement itself. It is assumed that the formal state-
ment has some unspecified constituents and, consequently, the formal
statement might be taken to be true for some values of these con-
stituents.
Assume that a set of formal statements K is given. If a subset M
of these statements is taken to be true, it defines a theory T over K.
Namely, theory T conjectures that only the statements in subset Mare true, the remaining statements being left unspecified.
Assume now that the formal statements in M are such that the
unspecified constituents represent formal objects. Formal statements
are then called proper statements. We have then:
Linguistic Definition: A general system is a set of proper statements.
2jThe second approach is based on set-theoretic considerations.
Formal objects are defined explicitly rather than as conceptual classes /via formal statements. (Conceptual class is understood in the sense of
Curry.2) One starts by defining a family of sets Xi " -, Xn. Each
set specifies a formal object. Namely, the formal object defined by
the set Xj can take on the form of any of the elements of the set
X;. The elements of Xi can be called values of the object in X;. The
cartesian product X of the family of sets is then formed
X = Xl * X2 * X * ... * Xn (1)
We can now introduce the following:
Explicit Definition: A general system is a proper subset of X.
X, c X (2)
Actually, a proper subset of the cartesian set X defines a relation
among the objects Xi, ..-, X". Notice that an element of the set Xf
-
8/19/2019 General System Theory.pdf
27/197
FOUNDATIONSFORA GENERALSYSTEMSTHEORY 7
can itself be a set, e.g., in the case of dynamic systems. The elementsof sets Xj are then time functions defined on appropriate intervals(tl, t2) and, therefore, they represent the infinite set of values which
the time functions will take in that interval.The explicit definition can be also stated in the following way:
A general system is a relation defined on the cartesian product X.
A general system is therefore specified by
X=X1*",*Xn and } (3)The difficulties in the linguistic as well as in the explicit specification
of a general system lie in the fact that both X and R can be infinitesets which can be specified only as inductive classes, i.e., by definingeffective processes which would generate them. The introduction ofsuch generating processes leads to the implicit definition of a generalsystem.
First, a systems term set Xj is specified by an effective process usingthe following prerequisites:
1. Set of initial elements.
2. Effective procedure for generating new elements of Xj from initialor previously generated elements.
3. Closure condition which requires that only elements generatedby (1) and (2) are elements of Xj.
Second, systems relations R are also specified inductively in thefollowing way:
1. A set of relations T is given:
T = {T¡, ..., T;} } (4)2. Any of the systems relations R; is represented as a sequence of
relations from the set T.
Ri - { Tin Tj" ..., Tjm;11 (5)
3. A set of rules of combination, P Pn}, is given whichspecifies how the sequences Ri = f Tl, are formed.
We can now introduce the following:
f Implicit (Syntactic) Definitions:A general system is defined by:``
1. A set of implicitly defined formal objects.2. A set of elementary transformations T.3. A set of rules P for forming the sequences of T.4. A set of statements indicating initial forms of the formal objects
for use in generating new forms of the objects.
-
8/19/2019 General System Theory.pdf
28/197
8 VIEWS ON GENERAL SYSTEMS THEORY
Notice that in the context of the linguistic approach items 1 through
3 define the class of formal statements and that only item 4 makes
possible the formulation of a theory of formal objects, i.e., a general
system.
OPEN AND CLOSED SYSTEMS
Consider an explicitly defined system. The subset Xs is essentially a
set of n tuples. A given n tuple, element of X8, will be called an exemplar.
The elements of these n tuples consist of all admissible values for the
formal objects. We now define a class of systems X,', X82, ... Xsmsuch that X8' c X8 for all j's.
A problem of primary importance in the application of systems theory
to real systems is related to the existence of effective means to distin-
guish among the systems X.1, ..., Xs"d. Assume that an exemplar
xi E X. is given. An equivalence relation, Ll(x2, Xs), defined on Xs
will permit the determination of a proper subset Xsi(L1) c Xs which
contains the given exemplar xi E X.i(L1). At the same time, the rela-
tion L1 determines which of thesystems
X. mare under con-
sideration, since only those systems that contain X,i(Ll) are considered
as being relevant. Assume now that a finite sequence of equivalence
relations Lit, L_ exists, so that the set X8i(L1, ..., L_) consists of a
single element. In other words, by applying the sequence of equivalence
relations L1, ..., the exemplar xi will be uniquely determined
since it will be the only element in the set X.i(L1, ..., Ln). Such a
sequence of relations is called an effective identification process. We can
now introduce the following concept:
r
A general system X$' is termed closed if and only if for every xi E x/
there exists an effective identification process.
By using appropriate effective processes it is possible to distinguish a
closed system X/ from any other system X 8 i eX.. The existence of
an effective process is not a property of a general system but rather the
result of the observer's relation with the system itself. An equivalence
relation Lj can be considered as a measurement while an effective proc-
ess (i.e., a sequence of relations L1, ..., Ln) can be considered as anexperiment performed on a real system.
\
A system for which the conditions outlined above are not satisfied,
i.e., for which there is at least one xi E Xsk for which an effective iden-
tification process does not exist, is called an open system.
In essence, if a system is open, it cannot be distinguished from some
other system X82. In general, a system becomes open if in the hypothesis
-
8/19/2019 General System Theory.pdf
29/197
FOUNDATIONSFORA GENERALSYSTEMSTHEORY 9
we are able to make about the system (and which we test by the ex-
periments), some of the principal constituents are omitted, e.g., asmaller number of formal objects is assumed.
Three typical cases of open systems are:1. A system which is not completely isolated from its environment
(the system has "disturbances" or uncertainties).2. A system which responds to the experimentation in such a way
as to cause permanent changes in its behavior (adaptive or self-
organizing systems).3. A system with which the experimenter is interacting; i.e., while
influencing the system he is, at the same time, influenced by it. (The
experimenter is "inside" rather than "outside" of the system duringexpérimentation. )
In dealing with an open system, one is confronted with a completelydifferent situation than when dealing with a closed system. Mostof the conceptual difficulties in decision making and control problemsarise from the fact that the system under consideration is open anddecision has to be made in the presence of uncertainties. The factthat some real systems result only in open general systems is of primary
epistemological importance.
STRUCTURE
The concept of the structure of a system is very important but has,as yet, received very little attention. In this section we introduce the
concept of the structure of a general system and show how it can beused in the
theoryof
implicitlydefined
systems.Consider an explicitly defined system. In general, the relation Rhas at least one constant constituent which has a specific value. Thus,R can be considered as a specific example of a more abstract relationwhich can be obtained when the given constituent is left unspecified.For example in the formal statement, "He is older than Jack," the sys-tems relation is, "older than." However, it can be considered as a spe-cial instant of the relation,
"is (of different age) than"or, equivalently, .
"is (older) than""(younger)
\ LIld
-
8/19/2019 General System Theory.pdf
30/197
O VIEWS ON GENERAL SYSTEMS THEORY
The system relation R can be considered as being defined by anabstract relation and the specific values of the unspecified constituents,so-called relational constituents;
R = {T,.n } (6)
where T = systems structure,= set of relationed constituents.
We now have the following definition:r _ ,
The structure of a system is obtained by abstracting the system,1/1i relation, i.e., by leaving unspecified all the relational constituents.
An interesting example of how to derive a system structure is offeredby the case of linear dynamic systems,.13 Consider a single variablelinear system described by
y(t)= f
t 00 F(T)x(1 - T) dT (7)
oryRx
The systems relation can be represented in at least two ways: First,by taking time function F(r) as the constituent so that the systemsrelation is
R = {C, F(T)} 1 (8)
where C is the algebraic operation of?convolution. The system struc-ture is then given by the operation C; T = C; t = F(r). Second, the
systems structure can be obtained by abstracting certain numerical
parameters associated with the time function F(T). For example, forthe first-order differential system the relation becomes
j 1 j(9)
ls J
where and + are operations of integration, multiplication, andaddition respectively, and constituents 0 and K are parameters of the
system called the time constant and gain respectively. Systemsstructure is then T = {1/s, ', +} and the relational constituents are
{0, K}.Recognition of a systems structure involves imbedding the systems
relation into a set of relations differing only in the values for someconstituents. If this set of relations is not explicitly given, one can
apparently select many different sets in which the system relations canbe imbedded. The selection of the set T is then a matter of practical
-
8/19/2019 General System Theory.pdf
31/197
FOUNDATIONS FOR A GENERAL SYSTEMS THEORY il 1
concern. For example, if one is concerned with modifying the systemso as to improve its behavior, the systems relation should be imbeddedin such a set that the selection of the relations can be accomplished by
selecting the value for the unspecified constituent. Such a selection canbe made in an organized manner, e.g., by using some method of optimi-zation. However, selection of the structure itself, i.e., the set of rela-tions, remains in the domain of a heuristic decision.
In the process of so-called systems identification, it is of primaryimportance to distinguish between the identification of the systemstructure and the identification of the system relations. Briefly, theproblem of identification is the following: Given a set of input-output
pairs,find the
system relation.The
solution of the identification prob-lem consists essentially of two steps: (1) the selection of the system'sstructure T and (2) matching of the system's behavior with the rela-tions in the set T in order to find the best values for the relational zconstituents t.
The second step can be undertaken systematically, but it is veryimportant to emphasize that it must always be preceded by the firststep which, in general, is based on guesses or experience. For example,in the case of a
dynamic system,the
assumptionthat the
systemis
linear is equivalent to the selection of the convolution operation forthe system's structure. The identification problem consists then ofselecting the time function F (T) in the best possible way.
Let us now show how the structure of an implicitly defined systemcan be found. As an appropriate representative of this class of systems,consider an elementary formal system (EFS) defined by Smullyan bythe following items: lo
1. An inductively defined class of finite sequences of strings whichare well-formed formulas defined on the three alphabets-for symbols,for variables, and for predicates.
2. Two binary operations-Ri for substitution, and R2 for modusponens.
3. A closure condition specifying that only elements obtained byrepeated application of 2 on 1 are admissible systems terms.
To find the structure of the EFS, we must exhibit its explicit defini-
tion. Two new concepts must be introduced:1. A set of designation signs 0 and t
0 = f 7*1> *, * > Tje * * * and l' = i3'u e2l (10)
0 is the set of designation signs associated with well-formed formulaswhile is the set of signs associated with each of the relations Il and R2.
-
8/19/2019 General System Theory.pdf
32/197
Z VIEWS ON GENERAL SYSTEMS THEORY
2. A new binary operation z = Mj[yj, rj] defined in the following way:
(a) (T 7"= Tj) - (z = 0)(1(11)
(b) (T -j) -' (z = yj)where yj is the result of the application of the relations Rl or R2 attime t = ij. Relation Mj is essentially a storing operation active ifT = Tj.
We can now introduce an explicit definition of the EFS, which wedenote as MEFS, by specifying:
A. An inductively defined class of well-formed formulas.B. A set E
having tripletsof the
designation signsas
elements,E = {ej}; ej = (p./, th p./) (12)
where p./ is either a designation sign Tior a well-formed formula ; p,j2hasthe same interpretation as gj'; is either or
C. Two triadic relations Rl and R2 and a set of binary operations{Mj}.
D. A closure condition.
The diagram of the system is given in Fig. 1. The designationtriplets of the input at any given time determine both the relation andthe formula on which the selected relation R1 or RZ will be applied.The result is "stored" by the relation Mj. By applying the appropriatesequence of designation triplets at the input, the systems theorems are
generated at the output.
y
-
mi .... * Mi Rltf1 R2 m
. x
FIG. 1. MEFS system,x = {Xi 1 ={
-
8/19/2019 General System Theory.pdf
33/197
....."., : .
FOUNDATIONSFORA GENERALSYSTEMSTHEORY 13
The structure of MEFS is now specified by the following set ofrelations:
Z' RI, R2, Mi, M2, "',Mn, ... } (13)
where T, in general, is an infinite set.Since EFS and MEFS are equivalent in the sense that both generate
the same set of theorems, T can be considered as the structure of EFS.The explicit definition of the system depends upon the type of the new
relation M. It is possible to derive the structure of the EFS by usingdifferent relations M/
zj = (14)where
Mlis defined in the
following way: z;(t;)
=
x(t; - Tj) ;i.e.,
the output of M;1 is the input delayed by Tj seconds. This representa-tion indicates the dynamic behavior of the system, but the program-ming of the input sequences of designation signs is much more complex.
The introduction of the concept of structure in the study of formal
systems permits an approach to decidability problems from the view-point of reproducibility discussed in the next section. Also, a systemsinterpretation of the Gbdel incompleteness theorems is possible. How-
ever, these problems are technically too complex to be dealt with
herein.
DECOMPOSITIONAND STATEOF THESYSTEM
An important (practical) problem concerns the possibility of de-composing or reticulating systems into subsystems. It is of interestto learn what can be said about this matter on the level of our present
general discussion. Let us first introduce the concept of decomposi-tion. Assume a system specified explicitly by a polyadic nth-orderrelation,
R[Xi, Xn] (15)The general decomposition procedure can be given in terms of a
relative product. A relation R is a relative product _of the relationsR1 and R2 if the following condition is satisfied:
(xRy) (zR2y)j (16)General decomposition procedure for the systems relation R would con-sist of finding two new relations R1 and R2 for which R is a relativeproduct. The system S is then reticulated into the subsystems
" '; X;, Z]R,[Xll ... 1 xi, ZI(17)
R2[Z, XHI, ...,
-
8/19/2019 General System Theory.pdf
34/197
14 VIEWS ON GENERAL SYSTEMS THEORY
Consider now the following problem: Given a system with a polyadicnth-order relation (eq. 16) what is the lowest order of the subsystemrelations into which this system can be decomposed? The result can
be presented in terms of the following theorem:
TheoremAn nth-order system can be:
1. Decomposed into (n - 2)-triadic relations En}.2. Decomposed into dyadic relations if and only if for every triadic
relation obtained from (1) the following conditions are satisfied:
(a) 1 (XiR/Zj) n Xi+2)]} (18)
(b) Zi = xi+l U Xi+2
Proof: The first part of the theorem can be easily proved by construc-tion. Let us represent the system relation R as a relative product ofR1 and Rz: "
R = RdR2
X2, X3)] n [R2(Xl, Zl, X4, ..., Xn)] (19)
Since a new term Zl is introduced with the new relations RI and R2,
no restrictions are imposed and the decomposition is possible.Next, Rz is represented as a new relative product
Rz = RalR4 (20)
One then has
[RI(Z¡, X2, X3)] n [Rg(Z2' Z¡, X4)] n [R4(Z2, Xl, X5, Xn)] (21)Proceeding in the same way, one obtains finally
Rzcx-1) = Rz(k-1)+1/Rz(k-1)+zi k = 1 ... (n - 3)
X2, X3)J n ZI, X4)]
n ... n [R2(n-4J+I(Zn-g, Zn-4, Xn-2)]
n [R2(.-4)+2(Z.-3, XI, Xn)] (22)
No restriction is imposed when introducing relative products and,therefore, the decomposition is possible. There are n - 2 triadic sub-systems in the system of eq. 22.
To prove the second part of the theorem, we consider a subsystemfrom eq. 22; i.e.,
yj3) (23)
-
8/19/2019 General System Theory.pdf
35/197
FOUNDATIONSFORA GENERALSYSTEMSTHEORY 15
where F/, Yj', and Y;3 are the (respective) terms. We write R; as arelative product,
Rj =
[Rs2(Zl, (24)
Assume that the condition from eq. 18 is satisfied, so that the medium
term is Yj'. One then has
yj2)]
= n (25)
Subsystem Ri is, therefore, decomposed into binary subsystems.
Since this is possible for every j, the system is decomposed into 2 (n - 2)binary subsystems. This shows sufficiency of the condition from thetheorem. To show that this condition is necessary, assume that there
exists one triadic subsystem for which the medium term cannot be oneof the three terms in the original relation. One would then have
Rj(Yj', Yj2, y_3) J
Ri = RulRi2 (26)
n y_2 Y/)]
Since Zl is a new term, the second relation is triadic. By introducingtwo new relations R;1 and R;, the order of relation has not been
reduced since Rj2 is again triadic. This completes the proof.
This theorem shows that, in general, a higher order system cannot
be decomposed into the subsystems with less than triadic relations.
Perhaps the most important application of this theorem is the
introduction of the concept of the state of a system. Consider asystem which maps a family of sets X2 into the set of elements Xl; i.e.,
X,RX2(t) (27)
where X2(t) indicates that the elements of the set are time functions,
i.e., the sets in themselves. Assume that the sets x2(t) eX2(t), are
finite and have p elements. Equation 27 is of the order p + 1 and,
according to the theorem, cannot be decomposed into subsystems with
order less than triadic. Assume that the elements of areordered.
22 () [X2(tl), 22 (2) , ' ' ' , X2(tp)] (28)
The system relation (eq. 27) is then
(29)
-
8/19/2019 General System Theory.pdf
36/197
166 VIEWSON GENERALSYSTEMSTHEORY
Consider the subset of all the elements of xz(t) with index higher than j,
x/et) = [22 (lj+1) , ' ' ' , X2(tn)] (30)
The system relation (eq. 27) is thenx2jr(t)] (31)
where x2jr(t) is the remaining subset
(32)
We now represent R as a relative product
XIR,[X2' (t),
Z'l
(33)
The term Xi now depends upon the medium term Z and does not
depend upon the elements of X2(t) having an index lower than j.Elements of Z represent now the state of the system. The decomposi-tion of eq. 33 can be understood to mean that Xi depends upon the stateof the system at t = tj and all future elements of x2, but does not de-
pend upon the prior elements, i.e., those having an index lower than j.
In the context of dynamic systems,this is
usually expressed by thestatement that xl does not depend on the values of xz(t) prior to the
time tj at which the state is known.This discussion can, of course, be easily extended to the case of in-
finite sets. For example, if we let the set x2(t) be of infinite cardinal-
ity, we can represent it as the two subsets; e.g.,
} (34)
with the conditionn c X21] n C X22] [i < i (35) j i
where is the universal quantifier.The system relation can then be written as
,X2 21 (36)
and, by introducing the relative product, one has
Z](37)
ZR(Xzz)
Elements of Z represent the state of the system.Notice that the concept of the state of a system is introduced as a
consequence of the theorem given in this section. Namely, the theorem
-
8/19/2019 General System Theory.pdf
37/197
FOUNDATIONSFORA GENERALSYSTEMSTHEORY 17
states that, if one wants to reduce the order of the system relation byeliminating dependence upon all the elements of the subset X22, thenew relation has to be at least triadic, except in the case where the
conditions of the theorem are satisfied. The three terms of the triadicrelation are, then, input, output, and state.
ATTRIBUTESAND BEHAVIOROF SYSTEMS
The similarity in the methodology of the sciences and general systemstheory, as mentioned in the first section, cannot be overemphasized.The
jobof a
general systemstheoretician
begins onlyafter a
generalsystem is given, e.g., in its explicit form R[Xi , ... , Xn]. To learn aboutthe system by considering every element in the set of n tuples X. isclearly impossible. The situation is analogous to the study of a realsystem without the guidance of an appropriate theory. An infinitenumber of experiments would then be required. For the understand-ing of the type of systems under consideration, it is necessary to dis-tinguish the system from some other, possibly similar, systems by meansof a relatively small number of statements. These statements should
refer to the system in its entirety, as specified by the explicit definition,and not to one particular n tuple.
Consider the cartesian set X = Xl * X2 * ... * Xn and construct a
propositional function L(x) with the individual variable x ranging overthe set X. If the proposition is true whenever x; E Xs, then L(x)is considered to define an attribute of a system.
An attribute of a system is a propositional function defined on Xand valid in X,,.
The behavior of a system is defined in the following way:
The behavior of a general system is a set of systems attributes
I (38)The important question is how to select a set of attributes
Lk which are of importance to a particular system and how toestablish their presence or absence. This is, of course, a difficult prob-
lem which ultimately has its origin in the domain of a real system,outside the formal studies of general systems. The solution dependson how the real system is to be used. However, a formal studyshould offer guidance in indicating the kinds of attributes that mightbe present in a given system. For example, when approaching the
study of any system, one would like to know whether the system is
-
8/19/2019 General System Theory.pdf
38/197
18 VIEWS ON GENERAL SYSTEMS THEORY
open or closed, what its structure is, how the state of the system can be
described, etc. There are other properties as, e.g., stability, which
might be of special interest in some problems but of minor importance
in others.Whenever one studies or talks about a general system, one uses a
language not included in the formal statements which form Xs. Sucha statement about the system is called a metastatement and the lan-
guage used is called a metalanguage. One can, of course, use the same
language for the metastatements as for the formal statements ofthe system, but the two types of statements should be clearly dis-
tinguished.
A convenient wayto
studya
general systemis to form the meta-
statements into a new system called a metasystem. A metasystemis selected in such a way that it properly reflects certain attributes ofthe system but is simpler to study than the original general system.
REPRODUCIBILITYAND CONTROLLABILITYOF SYSTEMS
As examples of attributes of a system which can be studied on the
general level, we introduce the concepts of controllability and repro-ducibility.
First we must introduce the concept of a performance functional.Consider an exemplar of the system behavior. System terms canthen be used to determine whether or not the particular behavior ofthe system :is acceptable. Technically, one introduces a mapping fromX. into the set of real numbers Q.
R T(X - Y) - Q (39)
The problem of controllability refers to the capability of achieving acertain performance by using a given system and a given set of inputs.The definition is the following:
A system is controllable in the set Qc if and only if for every qi E Q,there exists an xi E X such that
1 yi (40)where xi is the input of the system and yj is the output.
In other words, the specified set Qc is a subset of the set Q, that canbe generated by using all possible systems inputs.
Qc CQ.
-
8/19/2019 General System Theory.pdf
39/197
FOUNDATIONSFORA GENERALSYSTEMSTHEORY 19
The mapping T usually specifies an optimal point representing the best
performance which one could achieve if the system had no constraints.For example, if T is a quadratic functional, the optimal point is the
minimum of that functional. We denote the optimal point in q by qop,and introduce the following modified notion of controllability:
The system S is controllable if Q, = i q.pt,q opt- a!.The system is not controllable if for all x E X ; q > (qopt+ a).Another concept is related to the capability of the system to reproduce
certain outputs. This concept is called reproducibility and is introducedon a local basis by using dense sets. We have the following definition:
DefinitionSystem S is reproducible in the set Yr C Y. if there exists a subsetYrd C Yr so that:
1. Yrd is dense in Y,.2. For every yj E Yrdthere exists an Xj E X such that y; = i.e.,
3 x; lv E A'b) n E [yi = } (41)
Requirements for the density of the subset in Y in which the
reproducibility of the system is evaluated is introduced to avoid thepossibility of recognizing a system as reproducible if it can generateonly a set of isolated points in the output space Y,. Namely, for a
given multivariable system, one can always find some isolated points in
Y, for which the system relation and inputs will be just right so thatthe outputs can be reproduced. However, reproducibility has a full
meaning only if all the points which are close enough can be reproduced.A general method for testing reproducibility in this sense is based
on the following theorem:
Theorem
System S = YTX is nonreproducible in the subset F,.,: if:
1. There exist two relations Ri and R2 whose relative product is the
system relation T = RlIR2, i.e., [YTZ] H n (ZR2x».
2. The common term Z of the relative product T is a subset of the
space Y' that is of lower cardinality than Y, i.e.,
} 1 < "(42)(42)
F= 1
In other words,
[YTX] H { (YR1Y') fl (Y'R2X)l } (43)
-
8/19/2019 General System Theory.pdf
40/197
20 VIEWSON GENERALSYSTEMSTHEORY
Proof: Condition that an Ri exists denotes the existence of a mappingof the subset Y' into the set Y,
y' OE yDenote the subset of all points in Y which satisfies eq. 43 as V c Y.Consider a point in V :y E V, i.e., y is a reproducible point. Point yvis a vector yv = [yl(t), Let the vector space Y havethe metric which relates to the metric in the component spaces in the
following way:m
j=1To every element y'(t) = yi(t)J, eq. 43 uniquely defines anelement of the remaining set y"(t) = yn(t)]: Y" =
n Y']: y"(t) E Y". In other words, to every y'(t) correspondsa unique element y(t) or y"(t).
Consider now the neighborhood of the point yv = [y'(t), y"(t)];K[yv, e]. For any internal point ys = [y's, y"s] of the neighborhood K,one has
E (45)Assume that ys E V, i.e., satisfies, i.e., ys is a reproducible point. Thedistance function is now
mds = d[ym = L: yai(t)]
j=1
; (46)i m=
E y'sj(t)]+
E j=1 j=i-j-1
Consider now the point yz which has all the elements but one thesame as yv, the remaining element being the same as Clearly Y. isnot a reproducible point.
Y. = (47)
The distance function for yz is now
dz = Y.] = (48)It is apparent that
dz < ds (49)
This means that to every point Ysinside the neighborhood K(yv, e) there
corresponds a point closer to yv which does not belong to the set V. Inother words, no matter how small in a neighborhood one selects K(yv, e),
-
8/19/2019 General System Theory.pdf
41/197
FOUNDATIONSFORA GENERALSYSTEMSTHEORY 21
e > 0, there is always an inside point in K which does not belong to V.Since set V is its own closure V = V, this means that the set V isnowhere a dense set. This completes the proof.
Importance of the theorem stems from the fact that it offers a generalmethod for establishing non-reproducibility of a system. Since existenceof the relation YRI Y' is sufficient for a system to be non-reproduciblein a set Y, it suffices to exhibit the existence of such a relation.
The method for testing reproducibility can be now described in thefollowing way:
To show nonreproducibility of a system, it suces to derive startingfrom the system relations a relation among the outputs (or their prop-erties as
required bythe
metric)in which the
inputswill not be
present.In its generality, the derived method for testing reproducibility issimilar to the methods for testing different attributes of the systemsbehavior, e.g., Hurwitz criteria or Lyapunov functions, etc. The princi-pal weakness of the method is that it is only existential and does notindicate which technique should be used to construct the constrainingrelation Rl.
GOAL-SEEKINGBEHAVIOR
In the discussions thus far we have always considered a system as arelation among the terms without specifying the meaning which can beattached to these terms with respect to the systems behavior. Yet, aswas emphasized in the section entitled "Basic Characteristics ofthe Theory," it is the meaning given to various terms that dis-tinguishes general systems theory from a mathematical discipline.
Every term in a system description has a meaning relevant to thesystems behavior which is different from the meaning conventionallyassumed in mathematical studies of formal systems.
There are two principal avenues for conducting discourse about sys-tems and, consequently, two contexts in which to give meaning to sys-tem terms. These two contexts differ principally in the relation ofthe observer to the system. In the first approach, one considers the
system from the outside and defines its behavior in terms of the
mapping of one subset of terms (inputs and states) into another.This way of looking at the system is termed the terminal approach.In the second approach one has knowledge of certain invariable
aspects of the system which reflect its goal. Also, one is aware of theactivities of the system which are responsible for the goal seeking.This approach is consequently called the goal-seeking approach. The
-
8/19/2019 General System Theory.pdf
42/197
22 VIEWS ON GENERAL SYSTEMS THEORY
goal-seeking activities are again transformations among some specified
system terms, but the system attributes which specify the system's
behavior now refer to the particular meaning given to these transfor-
mations.In the discussions in the previous sections, the terminal approach
has been used. To develop the goal-seeking approach, the basic con-
cepts of decision-making, control, adaptation, self-organization, etc.,
must be introduced on a similar, definite basis as were the concepts for
the terminal approach. Time and space limitation do not allow us to
consider the details of the goal-seeking behavior here. We make only
the following general remarks: (1) Every system, at least in principle,
can be described either via a terminal approach or a goal-seeking ap-proach. It is the relation of the observer to the system and the knowl-
edge which he possesses about the system that determine the approach
to be used. There are, of course, systems (e.g., mechanical systems)
where the goals are either unknown or, to our best knowledge, non-
existent. In these cases the introduction of a goal-seeking description
is artificial or trivial. We do not feel that one should be too concerned
with such special cases, since, for most interesting systems, both
approaches have been successfully used. (2) The behavior of goal-seeking systems is much more complex than the behavior of terminal
systems. (3) The behavior of goal-seeking systems is much less
understood than the behavior of terminal systems.
If a system can be successfully described by either of the approaches,
terminal or goal-seeking, and if the former is simpler, then why do we
need the latter? The answer, essentially, lies in the possibility for using
the system description to predict the system's behavior. That is, if
one has the goal-seeking description of a system's behavior, one hasfar more powerful means of predicting the system behavior under con-
ditions which are different from those existing in the previously con-
ducted experiments.
Let us illustrate this by an example: Consider an animal sub-
jected to a pair of stimuli. The first stimulus, light, is applied a
short interval in advance of the second stimulus, electric shock. The
animal can escape from the area under electric voltage by crossing a
barrier. By repeating the experiments, a learning curve for the animalcan be obtained by plotting escape time versus number of experiments.
A typical learning curve is given in Fig. 2. Disregarding some tran-
sients, the learning curve has a sharp discontinuity. It occurs roughly
at the time when the animal discovers the relation between the light
-
8/19/2019 General System Theory.pdf
43/197
FOUNDATIONS FOR A GENERAL SYSTEMS THEORY 23
and the shock stimuli, and starts running toward the escape barrier
immediately after the light appears.
It is apparent that in order to describe the discontinuous learning
curve in Fig. 2, one cannot use the same stimuli-response transforma-tion obtained for the same animal under différent conditions, e.g.,
those resulting in a continuous learning curve. Therefore, if one wants
to explain the behavior of a learning system under différent conditions
one has to have better knowledge of the learning process inside the
system itself. In such a situation, a model based on the goal-seeking
approach is very powerful. For example, if one is able to discover the
method used by the system for making the decision to switch from
one mode of behavior describedby
theexponential
curve to the other
mode described by the horizontal line, one would be in a position to
describe the behavior of the system for a variety of conditions without
having to conduct specific experiments.
Some basic concepts for the case of simpler goal-seeking systems
have already been stated in full generality. However, goal-seeking
systems may be very complex and may have a large number of differ-
ent and conflicting goals as well as a corresponding number of goal-
seeking subsystems. Although some progress toward the formaliza-
tion of the theory for these systems has been made in an initial
_-
T
' . '<
.
1 1 1 1 1 y-Light"
I Shock
LightedArea
III.--JEscapeTime
\ \
\
I
f 1 f \
/
1 1 1 \
\
\
\
n0 .
Number of experiments
FIG. 2.
-
8/19/2019 General System Theory.pdf
44/197
24 VIEWS ON GENERAL SYSTEMS THEORY
study of the so-called multilevel systems, there is still considerablework to be done in this direction. At the same time, this might wellbe an avenue offering a great return for research effort on general sys-
tems theory.
REFERENCES
1. Bertalanfy, L., "An Outlineof GeneralSystems Theory," The British Journalof the Philosophy of Science, 1, No. 2 (1950).
2. Curry, H. B., Outlines of a Formalist Philosophy of Mathematics, North
Holland Co., Amsterdam, 1951.
3. Ashby, W. R., Introduction to. Cybernetics, John Wiley and Sons, New York,1955.
4. Boulding, K., "General Systems Theory-Skeleton of Science," General Sys-tems Yearbook, 1, University of Michigan, Ann Arbor (1956).
5. Zadeh, L., "From Circuit Theory to Systems Theory," Proceedings IRE, May(1962).
6. Mesarovié, M. D., and Eckman, D. P., "On Some Basic Concepts of a Gen-
eral Systems Theory," Proceedings of the Third International Conference on
Cybernetics, Namur, Belgium (1961).
7. Mesarovié, M. D., "Towards the Development of a General Systems Theory,"
Neue Technik (August 1963).8. Mesarovié, M. D., "On Self-Organizational Systems," in Self-Organizing Sys-
tems, Spartan Press, Washington, D. C., 1962.
9. Mesarovié, M. D., "Self-OrganizingControl Systems," in Symposium onDiscrete Adaptive Processes, AIEE (1962).
10. Smullyan, R. M., Theory of Formal Systems, Princeton University Press,
Princeton, N. J., 1961.
11. Mesarovié, M. D., General Systems Theory, Notes from the class lectures
delivered at Case Institute, 1960.
12.Curry,
H. B., Foundation of Mathematical Logic, MeGraw-Hill Book Co.,New York, 1963.
13. Mesarovié, M. D., The Control of Multivariable Systems, MIT Press and
John Wiley and Sons, New York, 1960.
-
8/19/2019 General System Theory.pdf
45/197
CHAPTER2
GENERAL SYSTEMS AS APOINT OF VIEW
KENNETH E. BOULDING
(A Small Cry of Distress, from a Not Very Mathematical Mnn)l'm like a rat within a maze,When faced with sigma's i's and j's,And problems soon become enigmasWhen wrapped in i's and j's and sigma's.
K. B
I teach a course in general systems as part of the Honors Program ofthe University of Michigan.* I get an excellent cross section of the
undergraduate seniors taking the Honors Program in a wide varietyof departments. After taking it one year, a student came to me and
said, "I haven't learned a thing in this course but I have got a new
point of view." 1 confess 1 was encouraged by this remark and feltthat the course had probably been justified. 1 hoped, of course, thatthe student had picked up a certain amount of information at leastabout some of the more exciting intellectual developments which are
going on in our day. I presume he must have learned something or he
would not have obtained a new point of view. I must confess, how-ever, that giving a new point of view rather than imparting informationas such is one of the main objectives of the course, which is exactlywhy I have called it "general systems."
I have implied above that general systems is a point of view ratherthan a body of doctrine. In the future it may develop into a body of
doctrine, for it is difficult to find any intellectual movement which doesnot. Perhaps, indeed, this is inevitable and indeed desirable. One
cannot remain forever perched on a point of view, however pleasingthe prospect, and one must go down and occupy the land. At themoment it would be presumptuous to claim that there is any clearlydefined body of theory which could be identified with the name "general
* It is listed in the University of Michigan catalog (1962/63) as College Honors498.
25
-
8/19/2019 General System Theory.pdf
46/197
26 VIEWS ON GENERAL SYSTEMS THEORY
systems." Nevertheless there is a general systems point of view and
there is what DeSolla Price has called "an invisible college," of people
who recognize each other as possessing this point of view. What 1
wish to do, therefore, is to try to identify some of the main elementsof the point of view even though this may represent the first slippery
step toward a body of doctrine and may, indeed, reveal the distressing
fact that there is not one point of view in general systems but a large
number.
Any point of view depends on certain value presuppositions and
positions of the viewer, for we all view the world from some high-
valued peak within our own welfare function. Things which we value
more highly loommore
closelyin our mental
landscape.This I call
the iron law of perspective. 1 know of no real escape from this iron
law. Even the attempt in the world of art to look at things from all
sides, as the cubist does, sometimes ends up by looking at them from
no side at all, and results in breaking up order into something peril-
ously akin to chaos. Even if our welfare function sets a high value
on the virtues of objectivity and a Cartesian clarity, we still view the
world from a perspective, for those things which are clearly quantifiable
andorderly
willoccupy
theforeground,
and we willrelegate
Celtic
twilights, mystical experience, and all things of clouded and brooding
significance to the jungle edges of our tight little intellectual clearing.
The only way to bend the iron law of perspective is to know as well
as we can the point from which we view. Then we can at least know
intellectually, even if not perhaps viscerally, that the looming fore-
ground is not as large nor the shadowy background as small as it seems
to us. Not even general systems, broad as its landscape claims to be,
can escape this iron law, and we must therefore look frankly at the
value presuppositions which are likely to lead to a general systems
point of view.
The first of these presuppositions is a prejudice in favor of system,
order, regularity, and nonrandomness (all these words being roughly
synonymous), and a prejudice against chaos and randomness. Along
with the poets, the general systems type of person has "rage for order,"
as Austin Warren has called it. He will certainly be fond of mathe-
matics, almost certainly fond of music, and he may have a half-
ashamed passion for the eighteenth century.
The next prejudice is a simple corollary from the first, that is, the
whole empirical world is more interesting ("good") when it is orderly.
It is to the orderly segments of the world, therefore, that the general
systems man is attracted. He loves regularity, his delight is in the
law, and a law to him is a path through the jungle.
-
8/19/2019 General System Theory.pdf
47/197
GENERALSYSTEMSAS A POINTOFVIEW 27
Now comes the main article of the general systems faith, for the first
two, after all, he shares with many scientific specialists; that is, theorder of the empirical world itself has order which might be called
order of the second degree. If he delights to find a law, he is ecstaticwhen he finds a law about laws. If laws in his eyes are good, lawsabout laws are simply delicious and are most praiseworthy objects ofsearch. The critic may perhaps argue that the hunger and thirstafter this order of the second degree is merely a passion for the fami-liar. The general systems man, he will say, is the sort who would bereminded of Pittsburgh even in the middle of Bangkok, simply becauseboth are cities and have streets with people in them. The critic (I
somehow visualize him as a historian in a high collar) has a passionnot so much for order and familiarity as for things that are peculiar,unique, strange, and disjoint. As a general systems man, 1 will visithim in his lonely eyrie, but even there I will probably be remindedof something-much to his annoyance. To avoid circumlocutions, letme call my general systems man a generalist and my high-collaredhistorian a particularist. The generalist rejoices when he sees, for
instance, that in all growth patterns there are significant common
elements, such as nucleation, structural adjustment in the parts of thesystem, diminishing returns to scale, and ogive curves. A particularistbrushes this aside and rejoices in the fact that the growth of the floweris so different from the growth of a crystal, or the growth of Romeso different from the growth of Athens.
Because of his rage for order, our generalist is likely to set a highvalue on quantification and mathematization, for these are greathelps in establishing order. There is, therefore, a prejudice in their
favor anda desire
touse them as far as
(oreven farther
than) theycan
usefully go. Even the most passionate generalist, however, is likelyto admit that there are elements of the empirical world, such as
aesthetics, love, literature, poetry, human relations, religion, etc., whichresist quantification and yet are orderly in their own bizarre fashion.The generalist, however, has a strong desire to discover a continuum,which he enjoys perceiving, between quantifiable and nonquantifiableorder. He sees poetry as a nonrandom sequence of words, music asa nonrandom
sequenceof notes, different in kind but not in form from
any nonrandom sequence of numbers.Order is always perceived as an abstraction from the complex flux
of reality. It is indeed usually seen as a relation among abstractions
themselves, such as numbers, lines, or spaces. The difference, perhaps,between the general systems man and the pure mathematician is that,whereas the mathematician is content with the mere perception and
-
8/19/2019 General System Theory.pdf
48/197
28 VIEWSON GENERALSYSTEMSTHEORY
demonstration of abstract order, the general systems man is interestedin looking for empirical referents of these systems and laws of abstract
order, for it is this ability to perceive the infinite particularity of the
empirical world as examples of an abstract order which gives thatworld its unity and, indeed, makes it in some sense "good." To
justify the ways of God to man is an important task, even for atheists,and it is done by tracing the golden threads of abstract order throughthe infinitely fragmented dark and light mosaic of the world of ex-
perience.Thus the mathematician is content with having discovered the ex-
quisite abstract order implied in the exponential function and the
elegant properties of the strange number e (2.718). The general sys-tems man seizes upon this as the expression of a general law of growthat a constant rate, of which over short periods there are innumerable
examples in the empirical world, whether this be the growth of a
crystal, of a living body, of an organization, of a whole economy, or
perhaps of the universe itself. The mathematician devises an equa-tion which describes a familiar ogive curve; the general systems man
regards this as a pattern which is repeated over and over again in
the empirical world, as we move from one equilibrium system toanother at a différent level. The mathematician sees the solution ofsimultaneous equations as a problem in abstract order; the generalsystems man perceives this as descriptive of the equilibrium of an
ecosystem or of a price system. The mathematician develops differ-ence equations or differential equations as expressions of a relationshipamong purely abstract concepts; the general systems man considersthem descriptive of a large class of dynamic processes, whether one
which keeps the planets in their courses, one which determines themovement of a falling body, or one which describes the movement ofan economic system through time. Not all mathematical equations or
relationships have empirical referents. 1 find it hard to conceive, for
instance, any real empirical referent to the series of prime numbers,although I have one student who is working on this problem. Whenan empirical referent can be found, however, there is great rejoicingin the general systems heaven.
The process of finding empirical referents to formal laws can easilytake either one of two possible directions. We may find some elegantrelationship in the world of abstract mathematics and then look aroundthe world of experience to see if we can find anything like it, or we
may patiently piece out a rough empirical order in the world of ex-
perience and then look to the abstract world of mathematics to codify,simplify it, and relate it to other laws. A good example of this phe-
-
8/19/2019 General System Theory.pdf
49/197
GENERALSYSTEMSAS A POINTOF VIEW 29
nomenon is "Zipf's law." Zipf was a Professor of German at Harvardwho conceived a passion for counting things and for plotting the fre-
quencies of organized distributions on double logarithmic paper. A
purely empirical law seemed to emerge which showed that almost allorganized distributions-e.g., the size of cities, the frequencies of
words, the intervals between notes in music, or the distribution ofincome-when plotted on double logarithmic paper turned out to bea straight line. Following this purely empirical observation, whichitself may be open to some questioning, several attempts have been
made, the most successful perhaps by Herbert Simon, to developmathematical models which will produce organized distributions hav-
ing this property. Going to the other extreme, we find whole branchesof mathematics which have been developed by pure mathematicianswithout even a thought of empirical referents which have later turnedout to be of enormous significance in exploring the more refinedaspects of the empirical world, especially in physics. Even in thesocial sciences, the mathematical theory of convex sets was developedlong before it turned out to have an important empirical referent inlinear programming. The value system of the general systems man is,
therefore,different from that of the
puremathematician. The famous
Cambridge toast, "Here is to pure mathematics, and may it never be
any damned good to anybody," has been reworded by him to "Here'sto pure mathematics, and may it soon be good for something" or"Here's to empirical regularity, may we soon find a mathematicalexcuse for it." Whatever his other virtues, the general systems manis irretrievably impure in his tastes and his fundamental value system.
Simply because any point of view implies perspective as we have
seen,and
perspectiveis illusion
interpretedto mean
reality,the inter-
pretation may break down. The near things may be perceived aslarge, and the far things as small in fact. A point of view, therefore,implies certain dangers of misperception, and the general systemspoint of view is no exception to this rule. The general systems man,if he is honest, must admit these dangers and be prepared to facethem. Some of the dangers are obvious and are avoided fairly easily;others are more subtle and require a highly sensitized perception.
An obvious danger frequently pointed out is that an interest in the
whole empirical world and the attempt to view this world as a wholelead to superficiality and dilettantism. The whole empirical worldin our days, at least, is far beyond the capacity of any one mind toknow. My wife has a standard formula, for which 1 endeavor tobe grateful, for deflating me whenever my general systems visionssoar too high. She says simply, "If you are going to be the great
-
8/19/2019 General System Theory.pdf
50/197
30 VIEWS ON GENERAL SYSTEMS THEORY
integrator you ought to know something." Even the most renaissanceof renaissance men in our days cannot hope to know more than a
very small fraction of what is known by somebody. The general
systems man, therefore, is constantly taking leaps in the dark, con-stantly jumping to conclusions on insufficient evidence, constantly, in
fact, making a fool of himself. Indeed, the willingness to make a foolof oneself should be a requirement for admission to the Society ofGeneral Systems Research, for this willingness is almost a prerequisiteto rapid learning.
One obvious safeguard against the worst forms of superficiality isa firm foundation of knowledge in at least one empirical discipline.
Before the general systems man takes off into the outer space of hisignorance, he ought at least to have a launching pad in some disciplinewhere he can reasonably claim to be an expert. He must have the
courage not merely to take refuge in a well-tended little plot of spe-cialized knowledge but he should also feel that his most important andmost secure contributions comes when he brings back and applies tohis own specialized fields the insights which have come to him duringhis aerial surveys of the whole empirical universe. The ideal general
systems man must be willing to talk nonsense outside his own fieldbut must be equally unwilling to talk nonsense inside it.A danger less readily perceived, but perhaps characteristic of the
whole scientific enterprise, arises because the "rage for order" leadsto the perception of order where, in fact, no order exists. Alex Bavelashas reported (orally) some experiments in which he has given his sub-
jects random sequences or patterns, for instance, sequences drawn froma table of random numbers, and asked them to deduce any law or
principlewhich
governsthe
sequence.His
subjectsalmost
invariablynot only see order in these random sequences but also vigorously de-fend the order which they have perceived when he challenges it, and
they become angry with him when he suggests that sequences are, in
fact, random. I once had a sad experience in teaching business cycleswhen 1 began by asking the class to plot random sequences on a graph,the sequences being derived, for instance, by throwing dice or bypulling numbers out of a hat. Such sequences, when plotted, have allthe
appearanceof time series of most economic
variables,and I con-
fess my objective was to shatter the naive faith that the studentmight have in the strictly nonrandom character of economic fluctua-tions. Unfortunately, my experiments backfired. One student, whowas something of a gambler, thought that he did, in fact, perceive alaw in the sequence of throws of dice and spent most of the semester
trying to find what it was. He did not, as far as I recall, get a very
-
8/19/2019 General System Theory.pdf
51/197
GENERALSYSTEMSAS A POINTOF VIEW 31
high grade in the course, and I never did find out whethe